Mesoscale simulation of spherulite growth during polymer crystallization by use of a cellular automaton D. Raabe * Max-Planck-Institut f€ ur Eisenforschung, Max-Planck-Str. 1, D€ usseldorf 40237, Germany Received 10 June 2003; received in revised form 26 January 2004; accepted 11 February 2004 Available online 16 March 2004 Abstract The paper introduces a 3D cellular automaton model for the spatial and crystallographic prediction of spherulite growth phe- nomena in polymers at the mesoscopic scale. The automaton is discrete in time, real space, and orientation space. The kinetics is formulated according to the Hoffman–Lauritzen secondary surface nucleation and growth theory for spherulite expansion. It is used to calculate the switching probability of each grid point as a function of its previous state and the state of the neighboring grid points. The actual switching decision is made by evaluating the local switching probability using a Monte Carlo step. The growth rule is scaled by the ratio of the local and the maximum interface energies, the local and maximum occurring Gibbs free energy of transformation, the local and maximum occurring temperature, and by the spacing of the grid points. The use of experimental input data provides a real time and space scale. Ó 2004 Acta materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Simulation; Texture; Polymer; Crystallization; Spherulite growth; Microstructure; Transformation; Avrami 1. Motivation for introducing a cellular automaton model for spherulite growth 1.1. Basics of cellular microstructure automata Cellular automata are algorithms that describe the spatial and temporal evolution of complex systems by applying local switching rules to the discrete cells of a regular lattice [1]. Each cell is characterized in terms of state variables which assume one out of a finite set of states (such as crystalline or amorphous), but continu- ous variable states are admissible as well (e.g., crystal orientation) [2]. The opening state of an automaton is defined by mapping the initial distribution of the values of the chosen state variables onto the lattice. The dy- namical evolution of the automaton takes place through the synchronous application of switching rules acting on the state of each cell. These rules determine the state of a lattice point as a function of its previous state and the state of the neighboring sites. The number, arrangement, and range of the neighbor sites used by the switching rule for calculating a state switch determines the range of the interaction and the local shape of the areas which evolve. After each discrete time interval the values of the state variables are updated for all lattice points in syn- chrony mapping the new (or unchanged) values assigned to them through the local rules. Owing to these features, cellular automata provide a discrete method of simu- lating the evolution of complex dynamical systems which contain large numbers of similar components on the basis of their local interactions. The basic rational of cellular automata is to try to describe the evolution of complex systems by simulating them on the basis of the elementary dynamics of the interacting constituents following simple generic rules. In other words, the cel- lular automaton approach pursues the goal to let the global complexity of dynamical systems emerge by the repeated interaction of local rules. The cellular automaton method presented in this paper is a tool for predicting microstructure, kinetics, and texture of crystallizing polymers. It is formulated on Acta Materialia 52 (2004) 2653–2664 www.actamat-journals.com * Tel.: +49-211-679-2278; fax: +49-211-679-2333. E-mail address: [email protected](D. Raabe). 1359-6454/$30.00 Ó 2004 Acta materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2004.02.013
12
Embed
Mesoscale simulation of spherulite growth during polymer crystallization by use of a cellular automaton
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Acta Materialia 52 (2004) 2653–2664
www.actamat-journals.com
Mesoscale simulation of spherulite growth duringpolymer crystallization by use of a cellular automaton
m3, r ¼ 1:18� 10�2 J/m2, re ¼ 9:0� 10�2 J/m2, a ¼4:55 �A, b ¼ 4:15 �A, ‘u ¼ 1:27� 10�10 m, nz ¼ 2000,
T1 ¼ 220:0 K, 1 ¼ 2:2� 10�12 kg s, as well as km ¼1 lm as length of one cubic cellular automaton cell.
These data give typical maximum growth velocities ofthe spherulite at the peak temperature between weak
secondary nucleation at large crystallization tempera-
tures and weak diffusion at low crystallization temper-
atures of the order of 10�7–10�6 m/s.
Polyethylene is a macromolecular solid in which the
molecular units are long chain-like molecules with C
atoms forming a zig-zag arrangement along the back-
bone and 2H atoms attached to each C. Fig. 2 gives aschematical presentation of the molecule and the unit
cell. The monomer unit of polyethylene is C2H4. The
crystals form by folding the chains alternately up and
down and arranging the straight segments between folds
into a periodic array. The crystal has orthorhombic
body centered symmetry (a 6¼ b 6¼ c, all angles 90�).Fig. 3 shows the growth velocities for a set of simula-
tions with different assumed temperatures for T1.Possible variations in the glass transition temperature
are of substantial interest for the simulation of spheru-
lite growth microstructures under complex boundary
conditions. The reason for this is that the glass transi-
tion temperature can be substantially altered by chang-
ing certain structural features or external process
parameters. Examples for structural aspects are the de-
pendence of the glass transition temperature on chainflexibility, stiffness, molecular weight, branching, or
crystallinity. Examples for external process aspects,
2660 D. Raabe / Acta Materialia 52 (2004) 2653–2664
which in part drastically influence the aforementioned
structural state of the material, are accumulated elastic–
plastic deformation, externally imposed shear rates, or
the hydrostatic pressure.
4.5. Nucleation criteria
Two phenomenological approaches were used in the
simulations to treat nucleation, namely site-saturated
heterogeneous nucleation and constant homogeneous
nucleation. The calculations with site-saturated nucle-
ation condition were based on the assumption of ex-
ternal heterogeneous nucleation sites such as provided att ¼ 0 s by small impurities or nucleating agent particles.
The calculations with constant thermal nucleation were
conducted using the nucleation rate equation of Hoff-
man et al. [28,30].
5. Simulation results and discussion
5.1. Kinetics and spherulite structure for site-saturated
conditions
This section presents some 3D simulation results for
spherulite growth under the assumption of site-saturated
nucleation conditions using a cellular automaton with
10 million lattice cells. These standard conditions are
chosen to study the reliability of the new method withrespect to the kinetic exponents (comparison with the
analytical Avrami–Johnson–Mehl–Kolmogorov solu-
tion), to lattice effects (spatial discreteness), to topology,
and to statistics.
Figs. 4 and 5 show the kinetics for simulations at 375
and 360 K, respectively. The calculations were con-
ducted with 0.05%, 0.01%, 0.005%, and 0.001% of all
cells as initial nucleation sites (site saturated). The fig-ures show the volume fractions occupied by spherulites
as a function of the isothermal heat treatment time. The
resulting topology of the spherulites is given in Fig. 6. It
Fig. 4. Avrami analysis of the volume fraction occupied by spherulites (fs) as375 K; 107 cells.
is important to note in this context that the depicted
spherulite volume fraction must not be confused with
the crystalline volume fraction, since the spherulites are
two-phase aggregates consisting of heavily branched
crystalline lamellae with amorphous chains betweenthem (Fig. 1). The kinetic anisotropy which is princi-
pally inherent in such structures at the nanoscopic scale
in terms of secondary nucleation events on the lateral
lamellae surfaces and of in-plane lamella growth [36,37]
is homogenized at the mesoscopic scale, where the
growing spherulites typically behave as isotropic
spheres, Eq. (1). In the present study this results in
Avrami-type kinetics with an exponent of 3.0� 0.1%.The small observed deviations of �0.1% from the ideal
kinetic exponent of 3.0 occur at small and at large times
due to the influence of the discreteness of the cubic au-
tomaton lattice at these early and late growth stages,
respectively. The scatter in the data for the spherulite
volumes at large times (particularly above 99% spheru-
lite volume) is even more important than that for short
times (see Figs. 4(b) and 5(b)). This effect is essentiallydue to the fact that the transformation from the amor-
phous to the spherulite state performed by the last cell
switches becomes increasingly discrete owing to the
changing ratio between the final non-transformed vol-
ume and the switched cells. The relevance of the data
predicted for the final stages of the transformation is,
therefore, somewhat overemphazised and should be
treated with some skepticism for instance when analyz-ing kinetical exponents at the end of the transformation.
It should also be mentioned that three simulation
runs were conducted for each set of starting conditions
in order to inspect statistical effects arising from the
Monte Carlo integration scheme, see Section 4.3 and
Eq. (14). Figs. 4 and 5 reveal that the statistical fluctu-
ations arising from this part of the simulation procedure
are very small (note the similarity of three curves on topof each other in both sets of figures).
The simulated spherulite growth kinetics are in good
qualitative accord with experimental data from the
a function of time; site-saturated nucleation conditions; polyethylene;
Fig. 5. Avrami analysis of the volume fraction occupied by spherulites (fs) as a function of time; site-saturated nucleation conditions; polyethylene;
360 K; 107 cells.
Fig. 6. Three subsequent sketches of the simulated spherulite microstructure at 375 K for 103 site-saturated nucleation sites (0.05% of 107 cells). The
gray scale indicates the respective rotation matrices h ¼ hðu1;/;u2Þ of the spherulite nuclei (not of the entire spherulites) expressed in terms of their
rotation angle relative to the sample reference system neglecting the rotation axis. The residual volume (white) is amorphous.
D. Raabe / Acta Materialia 52 (2004) 2653–2664 2661
literature [38–40]. It is important to underline in this
context though that the comparison of the experimen-
tally observed literature data [38–40] with the here
simulated crystallization kinetics remains at this stage
rather imprecise. This is due to the fact that some of the
simulation parameters as required in the current simu-
lations such as for instance the nucleation rates, the
cooling rates, and the exact structural data of the usedPE specimens were not known or not sufficiently docu-
mented in the corresponding publications, at least not in
the depth required for this type of simulation. In a
Fig. 7. Cahn–Hagel diagrams total interfacial area between the spherulitic ma
as a function of the spherulitic volume fraction; site-saturated nucleation co
second step it is, however, absolutely conceivable to
conduct real one-to-one comparisons between simula-
tion and experiment since some recent publications
provide more detailed experimental data (e.g., see
[41,42]).
Fig. 7 shows the corresponding Cahn–Hagel dia-
grams for the two simulations presented in Figs. 4 and 5.
Cahn–Hagel diagrams quantify the ratio of the totalinterfacial area of all spherulites with the residual
amorphous matrix and the sample volume as a function
of the spherulitic volume fraction. For an analytical
terial and the residual amorphous matrix divided by the sample volume
nditions; polyethylene; 107 cells. (a) 375 K and (b) 360 K.
2662 D. Raabe / Acta Materialia 52 (2004) 2653–2664
Avrami-type case and site-saturated conditions this
curve assumes a maximum at 50% spherulite growth
which is well fulfilled for the present simulations.
Fig. 8 shows the resulting spherulite size distributions
in terms of the spherulite volumes for the four cases0.001% cells as initial nuclei (a), 0.005% cells as initial
nuclei (b), 0.01% cells as initial nuclei (c), and 0.05%
cells as initial nuclei (d) (site-saturated conditions). The
diagrams use a logarithmic axis for the spherulite size
classes and a normalized axis for the spherulite fre-
quencies (number of spherulites in each size class di-
vided by the total number of spherulites). Such a
presentation provides a good comparability among thefour simulation sets. The results were fitted by using a
logarithmic normal distribution (solid line in each dia-
gram) which is usually fulfilled for Avrami-type growth
processes with site-saturated nucleation conditions. The
comparison shows that the simulations indeed repro-
duce the statistical topological behavior of Avrami
processes very well.
The data nicely document the gradual shift of thefinal spherulite size from conditions with a very small
number of initial nuclei (Fig. 8(a), 0.001% cells as initial
nuclei, large average spherulite size after heat treatment)
to conditions with a very large number of initial nuclei
(Fig. 8(d), 0.05% cells as initial nuclei, smaller average
spherulite size after heat treatment).
Fig. 8. Spherulite size distributions in terms of the spherulite volumes for fou
for the spherulite size classes and a normalized axis for the spherulite frequenc
of spherulites). The lines represent curve fits by use of a logarithmic normal
0.05% nuclei.
5.2. Kinetics and spherulite structure for constant nucle-
ation rate
The simulations were also conducted for constant
nucleation rate using a set of different activation ener-gies for nucleation. In the present study, this results in
Avrami-type kinetics with an exponent of 4.0� 1% (see
Fig. 9). This deviation of �1% from the ideal analytical
exponent of 4 represents a rather large scatter which,
however, can be attributed to the discreteness of the
cubic automaton lattice. The fact that the deviation in
kinetics (�1%) is much larger than that observed for the
simulations with site-saturated nucleation conditions(�0.1%) (Section 5.1, Figs. 4 and 5) can be explained by
the temporal change in the ratio between the remaining
amorphous matrix material which is not yet swept and
the new nucleation cells. This means that – since the
residual matrix volume which is not swept is becoming
smaller with each simulation step – each new nucleus
which is added to the lattice during one time step oc-
cupies an increasingly larger finite volume relative to therest of the material. The analytical result of 4, however,
is based on the assumption of a vanishing volume of new
nucleation sites. Fig. 10(a) shows three subsequent
microstructures of the same simulation. Fig. 10(b)
shows results form a corresponding set of simulation
results with site-saturated conditions, but under the
r different site-saturated nucleation conditions using a logarithmic axis
ies (number of spherulites in each size class divided by the total number